Rogue AI Agent Caught Typing Without Permission !
Years ago early in my career I worked in construction related business and Mike, this good ole Carolina boy, who worked for me as a job superintendent had a whole litany of colloquialism's he could throw down on almost any topic. We were talking about one of our other employees who was really good at his job but unfortunately that day, made some bad decision that resulted in some ugly consequences. To paraphase Mike - it only takes one "WTF" to wipeout a whole lot of "That was amazing".
Recently one of my best co-developers, pictured here above, made we situp straight in my chair. I am extremely cautious and I learned early on what I could and could not expect from AI assistance in code development. My approach was to adopted a "Don't Trust but Verify !" policy. Even so, I was not fully prepared for what I saw - in our coding chat there popped up a command to be executed ..../User/redacted/developer/..... Nobody on my team nor anywhere in my contacts is named redacted.
Foruntately, I was able to prevent the command from running but really ??? WTF ??
Obviously, to say this was disconcerting would be an understatment. I reported it to Kiro as a bug and we will see what they say.
Meanwhile after doing a little research I found out I am not alone. Feeling perhaps a little thankful actually....AWS has not been so lucky.
Read more here: https://x.com/AnishA_Moonka/status/2031434445102989379